Writing
- Gossip ProtocolsSpreading cluster state through peer-to-peer exchange
- Conflict-Free Replicated Data TypesIndependent updates that converge without coordination
- ReplicationMoving data between replicas
- Leader ElectionChoosing a leader when heartbeats stop
- Failure DetectorsDetecting failures within a cluster of nodes
- Kafka PartitioningDistributing messages into separate logs within a topic
- Connection PoolingReusing database connections to cut latency and resource cost
- ReconciliationRepairing bad data with scheduled jobs
- Batching (LLM Inference)Using the OpenRouter Batch API to save money, and the openrouter-batch crate to do it from Rust
- FanoutWhen service calls multiply downstream
- LLM Tool CallingA practical guide to model-driven APIs
- N+1 Query Problem (SQL)How repeated lookups become an N+1 query problem
- Sharding vs. PartitioningHow to split data and choose shard keys
- Multi-Region DataRouting, consistency, and failover across multiple regions
- Talking To My ComputerUsing voice input to work with software
- Durable ExecutionHow long-running work survives failures
- Optimistic LockingPreventing lost updates with version guards
- The Commit LogHow databases commit before flushing every data page
- Change Data CaptureTracing a Postgres update through a CDC pipeline
- Street MapsEmbedding OpenStreetMap in a static blog post
- ASCII AnimationBuilding ASCII animations for the web
- Really Good SoftwarePower tools I rely on
- Mermaid DiagramsEmbedding Mermaid diagrams on the web
- Kubernetes in the BasementBuilding a Kubernetes cluster in my basement